Overexpression of MFAP4 inhibits the proliferation, migration, and invasion of bladder cancer cells

Abstract

Background

The incidence of bladder cancer (BLCA) has steadily increased in recent years, and it is characterized by a high recurrence rate, often leading to poor patient outcomes. Cystoscopy and urinary cytology examinations have several limitations. Therefore, identifying biomarkers associated with the diagnosis and prognosis of BLCA is crucial.

Methods

A bladder cancer dataset was obtained from the Genome Expression Online (GEO) database. Differentially expressed genes (DEGs) were screened, and the selected genes were subjected to weighted gene coexpression network analysis (WGCNA), least absolute shrinkage and selection operator (LASSO) regression analysis, and protein‒protein interaction (PPI) network analysis to identify key genes involved in bladder cancer. The diagnostic value and prognostic significance of these key genes were validated using ROC curves, box plots, and survival curves. The infiltration levels of key genes in 28 immune cell types and their intercellular associations were analyzed. Molecular docking validated the binding affinity between MFAP4 and BLCA therapeutic drugs. MFAP4 expression in BLAC cell lines and SV-HUC-1 cells was detected via qPCR and WB analyses. MFAP4-overexpressing UM-UC-3 cells were generated using liposome-based transfection. Functional assays were performed on transfected UM-UC-3 cells. CCK8 assays were used to evaluate the effect of MFAP4 overexpression on doxorubicin sensitivity.

Results

This study revealed a total of 393 differentially expressed genes (DEGs), including 63 upregulated genes and 330 downregulated genes. Weighted gene coexpression network analysis (WGCNA) revealed seven coexpression modules, with the turquoise module showing the strongest correlation with bladder transitional cell carcinoma (BLCA). Intersection analysis between the differentially expressed genes and the turquoise module genes yielded 106 genes. Through LASSO regression analysis and protein‒protein interaction (PPI) network analysis, four key genes—COL6A1, COL16A1, CXCL12, and MFAP4—were selected. ROC curves, box plots, and survival curves revealed that MFAP4 had a high diagnostic value and was strongly associated with BLCA prognosis. Most cell types exhibited reduced levels of immune infiltration. Molecular docking analysis indicated that MFAP4 strongly binds to doxorubicin and gemcitabine, suggesting its potential as a drug target for BLCA. Compared with that in SV-HUC-1 cells, MFAP4 expression was significantly downregulated in BLCA cell lines. MFAP4 overexpression significantly inhibited the proliferation, migration, and invasion of UM-UC-3 cells, with both the IC50 values and cell survival rates being significantly lower than those of the other two groups.

Conclusion

The present study identified MFAP4 as a potential diagnostic and prognostic biomarker for bladder cancer (BLCA). MFAP4 is associated with proliferation, migration, and invasion of UM-UC-3 cells. MFAP4 overexpression enhances chemotherapy sensitivity in UM-UC-3 cells, potentially aiding in early diagnosis and providing novel therapeutic targets.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05056-3.

Introduction

Cancer is currently among the most significant threats to human health. As the ninth most prevalent malignant tumor globally, bladder cancer (BLCA) presents marked sex disparities in incidence, with rates significantly higher in males than in females [1, 2]. Histologically, more than 90% of BLCA originates from urothelial cells and is classified as urothelial carcinoma (UC). On the basis of tumor depth, BLCA can be classified into nonmuscle-invasive bladder cancer (NMIBC) and muscle-invasive bladder cancer (MIBC) [3]. Among patients with NMIBC who undergo transurethral resection of bladder tumor (TURBT), the recurrence risk is high, exceeding 50% within one year, and the five-year recurrence rate can reach 90%. MIBC is highly malignant, prone to distant metastasis, and has a poor prognosis [4].

The pathogenesis of bladder cancer is complex and involves the dysregulation of multiple genes and the disruption of several signaling pathways. Significant progress has been made in recent years regarding the underlying involved molecular mechanisms. Research has revealed that endoplasmic reticulum oxidoreductase 1 A (ERO1A) is highly expressed in bladder cancer tissues and cells. ERO1A activates the JAK-STAT signaling pathway and inhibits autophagy by binding to ALOX5 and suppressing its phosphorylation, ultimately promoting the proliferation, migration, and invasion of bladder cancer cells. This discovery offers a novel direction for the development of new molecular biomarkers for the diagnosis and treatment of bladder cancer [4]. Furthermore, long noncoding RNAs (lncRNAs) serve as key regulatory molecules that play dual regulatory roles in bladder cancer progression. For example, MIR205HG is epigenetically regulated by the p63 transcription factor, maintaining tumor cell sensitivity to interferon signaling by suppressing the excessive activation of interferon-stimulated genes, thereby exerting an antitumor effect, whereas loss of LEADR expression may contribute to the transition of bladder cancer to the muscle-invasive type and is associated with resistance to immunotherapy [5].

In addition to lncRNAs, abnormalities in proteins and signaling pathways also contribute to the development of bladder cancer: The menin protein encoded by the MEN1 gene is highly expressed in bladder cancer tissues and is closely associated with poor patient prognosis. Menin synergistically binds to the CTNNB1 gene promoter with the TFAP2C transcription factor, promoting β-catenin transcription, activating the Wnt/β-catenin signaling pathway, accelerating cell cycle progression, and driving bladder cancer cell proliferation. The BAY-155 menin-specific inhibitor suppresses bladder cancer progression by targeting this regulatory axis, providing a novel potential target for clinical treatment [6]. In the field of immunotherapy, the NFATC1 transcription factor regulates cytotoxic gene expression by directly binding to the PRF1 promoter and GZMB enhancer, with a significantly greater binding affinity in CD8⁺ T cells than in CD4⁺ T cells. Different therapeutic approaches, such as anti-PD-L1 therapy and chemotherapy, induce T cells to activate specific signaling pathways, including LCK, MHC-I, and CLEC, providing novel molecular mechanisms for optimizing immunotherapy strategies in bladder cancer [7].

Single-cell and spatial transcriptomics studies have revealed that the myoCAF_PLXDC1 subpopulation of peritumor cell-like myofibroblasts is enriched in malignant tissues and strongly correlated with poor patient prognosis. This subpopulation colocalizes with ITGA8⁺endothelial cells, participates in the development of tumor-specific vascular microenvironments, and is regulated by ligands (such as IFNG and TGFB1) and transcription factors (such as PRRX1 and MXD4), providing a novel potential target for targeted therapy in metastatic bladder cancer [8]. Lnc-TAF12-2:1 in urinary exosomes acts as a competitive endogenous RNA (ceRNA) to trap miR-7847-3p, thereby alleviating its suppression of the ASB12 target gene. This promotes the proliferation and migration of bladder cancer cells while inhibiting cell cycle arrest and apoptosis. This regulatory axis provides novel candidate biomarkers and therapeutic targets for noninvasive diagnosis and targeted therapy of bladder cancer [9]. Single-cell chromatin accessibility and DNA methylation analyses have revealed distinct origins among bladder cancer subtypes, with MIBC potentially arising from basal cells and NMIBC possibly originating from superficial intermediate cells or umbrella cells. The progressive evolution of TM4SF1-positive cancer cell subpopulations (TPCs) is closely associated with tumor invasiveness. Epigenetic regulation of differentially methylated regions (DMRs) and intercellular signaling pathways, such as the EGF, TGFβ/BMP, and Notch pathways, play pivotal roles in tumorigenesis and progression, offering novel epigenetic insights into bladder cancer pathogenesis [10]. Additionally, bladder cancers with extremely high tumor mutational burden (TMB) exhibit unique molecular characteristics, including high expression of immune checkpoint molecules (such as PD-L1, PD-L2, and PD-1), somatic mutations in cancer-related genes (such as TP53, TP63, and NOTCH3), and high hypoxia scores. These features provide a theoretical basis for combining immunotherapy with personalized treatments targeting p53 or hypoxia-related pathways [11].

Despite ongoing advances in understanding the molecular mechanisms of bladder cancer, clinical diagnosis remains fraught with challenges. Approximately 90% of patients present with painless gross hematuria at the time of diagnosis [12]. Currently, cystoscopy remains the gold standard for diagnosing BLCA [13]. Urine cytology is the only widely used urine biomarker detection method, but its sensitivity and specificity are limited because of interference from inflammatory conditions, such as urinary tract infections and urolithiasis [14–16]. Particularly for patients with NMIBC at extremely high risk of recurrence, regular cystoscopy is needed. However, repeated invasive procedures significantly decrease the quality of life of patients [17]. Therefore, developing highly reliable, noninvasive screening methods [18] to increase diagnostic efficiency and reduce unnecessary cystoscopy represents a major challenge. Unfortunately, existing urinary biomarkers generally lack sufficient sensitivity and have not yet gained widespread approval for clinical application. The development of accurate, effective noninvasive BLCA detection methods is crucial, and related research continues to actively advance. Against this backdrop, efficiently screening and identifying novel BLCA-associated biomarkers using bioinformatics methods offers a highly promising research direction for developing next-generation, highly sensitive diagnostic tools.

Weighted gene coexpression network analysis (WGCNA) is a systems biology method that describes intergene connectivity patterns, enabling the evaluation of relationships across different genomes and the identification of highly correlated gene clusters [19, 20]. The least absolute shrinkage and selection operator (LASSO) algorithm is a widely used feature selection method in bioinformatics that is capable of screening for more precise and effective feature information. In recent years, WGCNA and the LASSO algorithm have been extensively applied across multiple fields and play particularly significant roles in clinical practice [21, 22]. On this basis, the present study aimed to employ WGCNA and LASSO regression methods to identify and validate biomarkers and prognostic indicators for bladder cancer. This approach provides new theoretical foundations and technical support for early diagnosis, prognosis assessment, and optimization of treatment strategies for this disease.

Materials and methods

Data source

The following BLCA datasets obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) were integrated and analyzed in the present study: GSE7476 (BLCA = 9, control = 3), GSE13507 (BLCA = 188, control = 68), and GSE52519 (BLCA = 9, control = 3). After integration, a total of 206 BLCA samples and 74 normal samples were included, with normal samples serving as the control group throughout the study.

Cell culture

The 5637 and T24 bladder carcinoma cell lines were obtained from iCell Bioscience, Inc. (China), while the UM-UC-3 bladder carcinoma cell line and SV-HUC-1 human immortalized bladder epithelial cells were obtained from Procell (China). The specific culture protocols for each cell type are elaborated as follows: UM-UC-3 cells were cultured in MEM supplemented with nonessential amino acids (NEAAs), 10% fetal bovine serum (FBS) and a 1% penicillin‒streptomycin antibiotic mixture. T24 and 5637 cells were maintained in RPMI-1640 medium supplemented with serum and antibiotics at the same concentrations as specified above. SV-HUC-1 cells were cultured in Ham’s F-12 K medium supplemented with serum and antibiotics at the abovementioned concentrations. All the cell cultures were incubated in a humidified atmosphere containing 5% CO₂ at 37 °C.

Screening differentially expressed genes (DEGs) and enrichment analysis

We performed differential gene expression (DEG) analysis on standardized BLCA data using the limma package in R to compare gene expression differences between normal bladder tissue and BLCA tissue. DEGs were selected on the basis of |log₂FC| > 1 and p < 0.05, after which differential gene expression was visualized via volcano plots and heatmaps. All data analyses were conducted in R software (v4.1.3). Functional enrichment analysis of the DEGs was performed using the clusterProfiler package in R, encompassing GO analysis, KEGG pathway analysis, and GSEA. The results of the GO enrichment analysis covered three dimensions—biological process (BP), cellular component (CC), and molecular function (MF). KEGG pathway analysis revealed key signaling pathways on the basis of the significantly enriched genes. GSEA further elucidated the potential biological significance of gene expression data by comparing gene sets between the control and experimental groups [23]. The results were considered statistically significant when P < 0.05.

Identification of key genes

In the present study, the WGCNA package in R was used to construct a gene coexpression network between bladder cancer (BLCA) and normal bladder samples, with the goal of identifying gene modules highly associated with the disease. In brief, through network topology analysis and iterative adjustment and validation of the soft threshold parameter (β), we ultimately selected the minimal β value that yielded a scale-free fitting index (R²) approaching 0.90. Using this optimal β value, the gene correlation matrix was converted into a weighted adjacency matrix, which was further transformed into a topological overlap matrix (TOM). Module-enriched genes (MEs) are used to merge gene clusters with similar expression patterns, forming modules that represent highly co-expressed gene sets. Module‒phenotype associations were calculated by computing Pearson correlation coefficients between and clinical indicators, which were visualized as module‒phenotype association heatmaps. Modules significantly associated with BLCA (P < 0.05) were selected for subsequent analysis. From the modules obtained via WGCNA, genes with the highest phenotypic correlation were extracted and cross-analyzed with differentially expressed genes, yielding 106 candidate genes, which were further screened using a LASSO regression model, ultimately identifying 20 hub genes. A protein‒protein interaction (PPI) network for these hub genes was constructed on the basis of the STRING database [24], with an interaction score threshold of > 0.3. The PPI network was subsequently visualized and analyzed using Cytoscape 3.10.1 software. Key genes were definitively identified through cross-analysis of the following four algorithms: multicore correlation (MCC), multinode correlation (MNC), degree, and clustering coefficient.

Verification of key genes

To validate the reliability of the key genes, ROC curve analysis was performed to assess their correlation with survival prognosis in bladder cancer patients, and the area under the curve (AUC) was calculated. When the AUC was ≥ 0.8 and P was < 0.05, the gene model was considered to have good predictive efficacy. Survival analysis and gene expression‒stage correlation plots were generated using the GEPIA database (http://gepia.cancer-pku.cn/index.html), with data explicitly derived from the TCGA-BLCA (The Cancer Genome Atlas-Bladder Urothelial Carcinoma) cohort—the core bladder cancer dataset integrated in GEPIA. A total of 404 bladder cancer tissue samples and 19 adjacent normal bladder tissue samples were included in the analysis, among which all 404 bladder cancer samples had complete gene expression profiles and corresponding overall survival (OS) data. The clinical characteristics of this cohort include age, gender, tumor stage, grade, and follow-up survival time. Box plots were constructed to compare key gene expression differences between normal bladder tissue and bladder cancer tissue, with P ≤ 0.05 indicating statistical significance. Additionally, subcellular localization information for key genes was retrieved from the GeneCards database (https://www.genecards.org/) to further elucidate their intracellular distribution characteristics and localization patterns, laying the foundation for gene functional validation.

Correlation analysis between signature genes and immune infiltration

Using cluster analysis and single-sample gene set enrichment analysis (ssGSEA), we quantified the infiltration levels of 28 distinct immune cell types in BLCA tissues and normal bladder tissues, generating heatmaps and violin plots. We further investigated the correlation between 20 key genes and immune cell types.

Molecular docking

To investigate the interactions between core gene-encoded proteins and commonly used chemotherapy drugs for bladder cancer (BLCA), the present study employed molecular docking techniques to analyze the binding affinity of doxorubicin (DOX), gemcitabine, and cisplatin [25] for core gene-encoded proteins. Molecular docking analysis was performed using the CB-Dock2 online tool [26] (https://cadd.labshare.cn/cb-dock2/php/index.php), a state-of-the-art protein-ligand blind docking platform that integrates structure-based cavity detection and template-based docking (powered by FitDock) with a dual-pipeline scoring strategy. The core scoring function of CB-Dock2 inherits from AutoDock Vina (v1.1.2). This hybrid force field scoring system quantifies binding energy by integrating multiple key energy terms, including: repulsive steric hindrance, hydrophobic interactions, and hydrogen bonding. For template-based docking workflows, CB-Dock2 first employs the PC-score (a three-dimensional molecular similarity score integrating atomic distances, mass differences, and bond-type similarity) to screen for optimal initial ligand conformations by aligning with homologous templates. Subsequently, the aforementioned AutoDock Vina scoring function is applied for energy refinement, ensuring consistency and accuracy in binding energy assessment across different workflows [26, 27]. Generally, Vina score < -5.0 kcal/mol indicates stronger binding affinity between two molecules in this scoring system.

Quantitative polymerase chain reaction (q-PCR) and western blot (WB) analyses

Total RNA was isolated from 5637, T24, UM-UC-3, and SV-HUC-1 cells using TRIzol reagent following the manufacturer’s instructions. Complementary DNA (cDNA) was synthesized utilizing reverse transcription reagents (Vazyme Biotech, China), and qPCR was performed to assess MFAP4 expression. The reaction mixture consisted of 10 µl of 2× SYBR Premix, 1 µl of cDNA template, 0.4 µl each of forward and reverse primers, and 8.2 µl of double-distilled water (ddH₂O). The qPCR cycling conditions were as follows: initial denaturation at 95 °C for 5 min, followed by 40 cycles of denaturation at 95 °C for 30 s and annealing at 60 °C for 30 s. The primer sequences (Sangon Biotech, China) were as follows: MFAP4 forward, 5′-CTTTGAGAACAACACGGCCTA-3′; MFAP4 reverse, 5′-TCCCGGTCGAAGGTAGAGAAC-3′;

GAPDH forward, 5′-GGAGCGAGATCCCTCCAAAAT-3′; and GAPDH reverse, 5′-GGCTGTTGTCATACTTCTCATG-3′. The 2^⁻ΔΔCt method was used to quantify relative MFAP4 mRNA expression levels.

Total protein was extracted using RIPA lysis buffer, and the protein concentration was determined by a bicinchoninic acid (BCA) assay. After the addition of 1× SDS‒PAGE loading buffer, the samples were denatured in boiling water (100 °C) for 10 min. Equal amounts (30 µg) of protein samples were loaded onto SDS‒PAGE gels and subjected to electrophoresis at a constant voltage of 120 V for approximately 1 h, followed by transfer onto polyvinylidene fluoride (PVDF) membranes. The membranes were blocked with bovine serum albumin (BSA) for 1 h and then incubated overnight at 4 °C with primary antibodies against MFAP4 (Proteintech, China) and β-tubulin (Beyotime Biotechnology, China). The membranes were subsequently incubated with fluorescent-conjugated secondary antibodies (Beyotime Biotechnology, China) at room temperature prior to data acquisition. β-Tubulin served as the internal control. Statistical analyses were performed using paired t tests, and densitometric analysis of the bands was conducted using ImageJ software.

Cell transfection

UM-UC-3 bladder cancer cells with the lowest relative expression of MFAP4 mRNA were selected for transient transfection with overexpression vectors. Cells were divided into the following three groups: UM-UC-3 group (control, untreated), NC group (transfected with a nonspecific sequence plasmid), and experimental group (transfected with a plasmid containing the MFAP4 target sequence). The plasmids used in the present study were synthesized by Weihui Biotechnology. The transfection efficiency was subsequently validated by qPCR and WB analyses, and the transfected cells were used for subsequent experiments.

Cell plate cloning, migration, and invasion assays

Cells from the UM-UC-3, NC, and MFAP4 overexpression group were seeded at a density of 5 × 10⁵cells/well into six-well plates and incubated at 37 ℃ in a 5% CO₂ incubator for 2 weeks. When more than 50 visible clones formed, the cells were washed 2–3 times with PBS and fixed with fixative for 15 min. After discarding the fixative, the cells were washed once more with PBS, and the cells were stained with crystal violet solution for 20 min. After the cells were rinsed with PBS, they were allowed to air dry at room temperature and photographed for documentation. After the three groups of cells were seeded into 6-well plates and confluency was obtained, a scratch assay was performed using a 200 µL pipette tip. The cell debris was removed with PBS, and serum-free medium was added. The cells were observed and photographed under an inverted microscope at 0 h, 24 h, and 48 h. The scratch healing rate was analyzed using ImageJ software. To assess the invasive potential of the three cell lines, the cells were digested, resuspended in serum-free medium, and seeded into the upper chamber of Transwell filters. The lower chamber contained medium supplemented with 10% FBS. After 24 h, the medium was aspirated, and the cells were fixed for 15 min, stained with crystal violet, washed with PBS, dried, and photographed under a microscope for counting.

Cell counting Kit-8 (CCK-8)

Chemotherapy sensitivity experiments were conducted using doxorubicin, with the highest MFAP4 score selected through molecular docking. UM-UC-3 cells, NC cells, and MFAP4-overexpressing cells were seeded at a density of 5 × 10³ cells/well in 96-well plates. The cells were cultured with 100 µL of complete medium containing doxorubicin at varying concentrations (0–64 µg/mL), with 6 duplicate wells per group. After incubation for 24 h, 10 µL of CCK-8 solution was added to each well, followed by an additional incubation for 2 h. The absorbance values were measured and recorded, after which the corresponding IC50 was calculated.

Statistical analysis

Bioinformatics analysis was performed using R software. Data analysis, statistical calculations, and graphing were conducted with GraphPad Prism. Paired t tests were used for comparisons between groups. The results with p values < 0.05 were considered to indicate statistical significance.

Results

DEGs and their functional enrichment analysis

On the basis of the GEO database, the present study retrospectively analyzed clinical data from 206 bladder cancer samples and 74 normal bladder samples across three GEO datasets. A total of 393 differentially expressed genes (DEGs) were identified with P < 0.05, comprising 63 upregulated genes and 330 downregulated genes. To visually represent the expression patterns of the DEGs, a volcano plot (Fig. 1A) was constructed to display the expression trends of all the genes. Concurrently, a heatmap illustrated the expression pattern clustering of the 393 DEGs between BLCA and normal bladder tissues (Fig. 1B). The results revealed statistically significant differences in gene expression patterns between BLCA and normal bladder samples, further validating the reliability of the DEGs. To explore the association of DEGs with biological processes, signaling pathways, gene sets, and sample phenotypes, Gene Ontology (GO) enrichment analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, and gene set enrichment analysis (GSEA) were performed. GO enrichment analysis revealed “collagen-containing extracellular matrix” as the most significantly enriched term, with high enrichment levels also observed for “muscle system processes,” “extracellular matrix organization,” “extracellular structure organization,” and “external envelope structure organization” (Fig. 1C). KEGG pathway analysis revealed that the differentially expressed genes were enriched primarily in the following pathways: “muscle cell cytoskeleton,” “adhesion complexes,” “rheumatoid arthritis,” “vascular smooth muscle contraction,” “cell cycle,” and “human T-cell leukemia virus type 1 infection” (Fig. 1D). Compared with those in normal bladder tissue samples, GSEA (Fig. 1E) revealed that genes in the bladder cancer (BLCA) group were significantly enriched in pathways related to effector CD8⁺ T cells, CD4 memory T cells, and regulatory T cells (Tregs). These findings suggested that biological processes associated with these immune cells may play critical roles in the development and progression of bladder cancer.

Fig. 1.

Differentially expressed gene analysis and enrichment analysis. A Volcano plot. B Heatmap. C GO pathway enrichment analysis of the DEGs. D KEGG pathway enrichment analysis of the DEGs. E Classic maps for gene set enrichment analysis (GSEA)

Identification of key genes

A coexpression network was constructed for all genes across the GSE7476, GSE13507, and GSE52519 datasets. During the analysis, the gene importance filter threshold was set to 0.5, and the gene‒module correlation filter threshold was set to 0.8, yielding seven modules with significant coexpression patterns and their core gene sets. Among these genes, the turquoise module (110 genes) showed the strongest association with BLCA (cor = 0.56; p = 2e-24), suggesting that genes in this module may play a key synergistic role in the pathological process of BLCA. Therefore, subsequent analyses focused on genes within the turquoise module (Fig. 2). To further refine the core genes, an intersection analysis was performed between the 110 genes in the turquoise module and the 293 DEGs identified earlier, yielding 106 genes. Internal cross-validation using the LASSO regression model on the 106 genes revealed 20 hub genes (Fig. 3A–C). To better understand the interrelationships, a protein‒protein interaction (PPI) network diagram for the 20 LASSO-selected hub genes was constructed (Fig. 3D). Using five algorithms in Cytoscape software (MNC, MCC, degree, and clustering coefficient), the top five node genes were selected as hub genes. Finally, cross-validation of multiple algorithms revealed the following four key genes: MFAP4, COL6A1, COL16A1, and CXCL12 (Fig. 3E).

Fig. 2.

WGCNA network construction and significant module identification. A Clustering dendrogram. B Module–trait association plot. C Cross-module gene significance analysis plot

Fig. 3.

Identification of key genes. A Intersection of genes between DEGs and WGCNA modules. B Selection of the optimal parameter (λ) for the LASSO model. C LASSO model. D PPI network. E Intersection of genes from the four algorithms

Validation of COL6A1, COL16A1, CXCL12, and MFAP4

In the present study, the gene expression levels of four selected key genes (COL6A1, COL16A1, CXCL12, and MFAP4) were analyzed in bladder cancer (BLCA) and normal samples. Compared with those in normal bladder samples, the expression levels of all four genes were significantly decreased in BLCA samples (Fig. 4A). Additionally, ROC curves (Fig. 4B) were plotted to evaluate the diagnostic value of the four pivotal genes in bladder cancer, with AUC values of 0.831, 0.881, 0.850, and 0.905 for COL6A1, COL16A1, CXCL12, and 0.905, respectively. These findings indicate that MFAP4 has high predictive value for BLCA patients and may serve as a potential prognostic biomarker. Violin plots revealed that the median MFAP4 expression level progressively increased with clinical stage and closely correlated with disease progression (Fig. 4C). In addition, Kaplan–Meier survival curves revealed significantly lower overall survival in the high-MFAP4 expression group than in the low-MFAP4 expression group (log-rank p = 0.0047) (Fig. 4D). Compared with patients in the low-expression group, patients in the high-expression group had a 1.5-fold increased risk of death (HR(high) = 1.5, p(HR) = 0.0052), suggesting that high MFAP4 expression is closely associated with poor prognosis in bladder cancer patients. Finally, subcellular localization analysis of these four key genes via the GeneCards website (https://www.genecards.org/) (Fig. 4E) indicated that all the genes were distributed extracellularly.

Fig. 4.

A Expression patterns of the four key genes in BLCA samples versus normal samples. B ROC curves for the four key genes. C Violin plot showing the relationship between BLCA stage and gene expression level. D Kaplan‒Meier survival curve. E Subcellular localization analysis of the four key genes. ***p < 0.001

Immune infiltration analysis of key genes

We generated an immune cell heatmap using ssGSEA to assess immune cell infiltration in BLCA and normal bladder tissue samples (Fig. 5A). The immune cell infiltration levels were significantly lower in BLCA tissue samples than in normal tissue samples. The violin plot (Fig. 5B) revealed that compared with normal bladder tissue samples, bladder cancer tissue exhibited reduced levels of activated B cells, activated CD8⁺ T cells, activated dendritic cells, eosinophils, immature B cells, immature dendritic cells, myeloid-derived suppressor cells (MDSCs), macrophages, mast cells, monocytes, natural killer cells, natural killer T cells, plasmacytoid dendritic cells, regulatory T cells, follicular helper T cells, type 1 helper T cells, effector memory CD4⁺T cells, memory B cells, central memory CD4⁺T cells, central memory CD8⁺T cells, and effector memory CD8⁺T cells but increased infiltration of CD56-high natural killer cells. Furthermore, correlation analysis confirmed that the infiltration levels of multiple immune cells were positively correlated with MFAP4 expression (Fig. 5C).

Fig. 5.

Analysis of immune infiltration in bladder cancer. Heatmap (A) and violin plot (B) showing the distribution of 28 immune cell types in normal versus bladder cancer samples. C Correlation analysis between immune cell infiltration and the expression of 20 core genes. *p < 0.05, **p < 0.01, and ***p < 0.001

Docking of MFAP4 with chemotherapy drug molecules

Molecular docking of three drugs—doxorubicin, cisplatin, and gemcitabine—with MFAP4 was performed to evaluate their binding affinities. The Vina scores for gemcitabine, doxorubicin, and cisplatin with MFAP4 were − 6.1 kcal/mol (Fig. 6A), -9.2 kcal/mol (Fig. 6B), and − 5.0 kcal/mol (Fig. 6C), respectively. These results indicate that MFAP4 strongly binds to common chemotherapy drugs used for bladder cancer, suggesting that MFAP4 may serve as a potential therapeutic target for bladder cancer treatment.

Fig. 6.

Molecular docking of MFAP4 with three common bladder cancer chemotherapy drugs. A Gemcitabine. B Doxorubicin. C Cisplatin

Verification of the transfection efficiency of MFAP4 overexpression

qPCR and WB analyses were employed to determine the relative expression levels of MFAP4 in 5637, T24, UM-UC-3, and SV-HUC-1 cells. Consistent with previous findings, the MFAP4 expression levels in all BLCA cell lines were significantly lower than those in normal human bladder epithelial cells (Fig. 7A and B). These findings suggested that MFAP4 may serve as a novel diagnostic marker for bladder cancer.

Fig. 7.

MFAP4 expression levels. A Expression levels of MFAP4 mRNA. B Expression levels of MFAP4 protein. C, D UM-UC-3 cells, which had the lowest MFAP4 expression levels, were transfected with the MFAP4 overexpression vector, after which the transfection efficiency was validated. **p < 0.01, ***p < 0.001, and ****p < 0.0001

Verification of MFAP4 overexpression transfection efficiency

qPCR and WB analyses revealed that MFAP4 overexpression was significantly greater in UM-UC-3 cells than in the NC group (P < 0.001) (Fig. 7C and D), confirming successful construction of the MFAP4 overexpression vector.

Overexpression of the MFAP4 gene inhibits proliferation, migration, and invasion of UM-UC-3 cells

To evaluate the proliferation, migration, and invasion capabilities of UM-UC-3 cells following MFAP4 overexpression, we conducted functional assays. A colony formation assay revealed that compared with both the UM-UC-3 group and the NC group, the MFAP4 overexpression group exhibited significantly lower colony formation rates (Fig. 8A). A wound-healing assay demonstrated that MFAP4 overexpression significantly inhibited UM-UC-3 cell migration (Fig. 8B). Further, the Transwell assay revealed that MFAP4 overexpression markedly suppressed UM-UC-3 cell invasion (Fig. 8C).

Fig. 8.

Overexpression of the MFAP4 gene inhibits the proliferation, migration, and invasion of BLCA cells. A The clonogenic assay revealed that the proliferation capacity of MFAP4-overexpressing UM-UC-3 cells was reduced. B The cell scratch assay revealed increased scratch width percentage and decreased migration rate in MFAP4-overexpressing UM-UC-3 cells at 24 and 48 h. C Transwell assays revealed reduced invasion capacity in MFAP4-overexpressing UM-UC-3 cells. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001

Effect of UM-UC-3 cell overexpression on sensitivity to doxorubicin (DOX)

Compared with those in the other two groups, the IC50 and cell survival rates in the MFAP4 overexpression group were significantly lower (P < 0.05). These findings indicate that MFAP4 overexpression increases the sensitivity of UM-UC-3 cells to DOX (Fig. 9).

Fig. 9.

Detection of UM-UC-3 cell sensitivity to doxorubicin. The cells were divided into UM-UC-3, NC, and experimental groups. The cell viability rates were compared among the three groups

Discussion

Bladder cancer is among the most commonly diagnosed malignancies of the genitourinary system [28], and it is characterized by high recurrence and mortality rates. According to 2022 GLOBOCAN statistics, more than 600,000 new cases of bladder cancer are diagnosed globally each year, with approximately 200,000 deaths, and both figures show a persistent increasing trend [29]. Smoking history is the primary risk factor for BLCA [30, 31], followed by occupational exposure (e.g., prolonged contact with chemicals, such as paint) and drinking water contamination [29]. Additionally, chronic urinary tract irritation or recurrent urinary tract infections caused by long-term catheterization are associated with increased BLCA risk [32]. The most effective current treatment combines surgery with chemotherapy [33], with the most potent chemotherapeutic agents being methotrexate, vincristine, doxorubicin, and cisplatin [25]. The optimal surgical approach is transurethral resection of bladder tumor (TURBT). Furthermore, for high-risk NMIBC, the gold standard is IVe immunotherapy [34]. Although some BLCA patients experience disease improvement following standard clinical treatment, the complex pathogenesis and molecular therapeutic targets remain incompletely understood. Consequently, identifying novel and promising diagnostic biomarkers and molecular therapeutic targets remains an urgent and challenging endeavor [35, 36].

In recent years, multiple biomarkers for bladder cancer (BLCA) have been identified. For example, Chi et al. reported that TEAD4 is highly expressed in BLCA cells and promotes epithelial‒mesenchymal transition (EMT) in cancer cells, serving as a key regulator of EMT in BLCA cells [37]. Jin et al. demonstrated that LAMA2 and RUNX2 promote proliferation, migration, and invasion in BLCA cells while inhibiting apoptosis [38]. Zhao et al. revealed that CDK6 plays a crucial role in the immune microenvironment of BLCA and influences tumor biological behavior by regulating tumor immune status, metabolism, and cell cycle-related signaling pathways [39].

In the present study, the WGCNA method was used to construct a coexpression network and investigate key genes associated with BLCA. In total, 393 bladder cancer-specific DEGs were identified, comprising 63 upregulated genes and 330 downregulated genes. We subsequently performed KEGG and GO enrichment analyses alongside GSEA on the 393 DEGs. GO enrichment analysis revealed significant enrichment in collagen-containing extracellular matrix and structural components of the extracellular matrix. KEGG enrichment analysis highlighted the cytoskeleton in muscle cells. Additionally, GSEA enrichment analysis revealed significant enrichment of effector CD8 + T cells, CD4 memory T cells, and Tregs.

WGCNA is widely used to identify gene modules strongly associated with diseases and to help screen disease-related targets. Using the WGCNA method, seven modules were generated, among which the turquoise module was strongly correlated with BLCA and contained 110 genes. The DEGs were intersected with genes from the WGCNA turquoise module, yielding 106 genes. LASSO regression analysis was subsequently performed on these genes to obtain 20 genes. A PPI network was constructed using the STRING database, yielding 10 hub genes. Finally, four key genes were identified through the intersection of the four algorithms. Biomarkers have been screened via WGCNA and LASSO regression analysis for numerous diseases in recent years. Cui et al. employed WGCNA to identify modules associated with type 2 diabetic cardiomyopathy and used LASSO regression to select six key genes as diagnostic and predictive biomarkers for this condition [40]. Chen et al. performed WGCNA, LASSO regression, and ROC analysis on data and identified three genes—ZNF695, CHEK1, and C15ORF42—that are highly expressed in MYCN-positive neuroblastoma (NB) tissues and cell lines, and they reported that these three genes are closely associated with prognosis in pediatric NB patients [41].

In the present study, analysis of the expression of genes in BLCA and normal samples revealed consistent and significant differences in the expression patterns of COL6A1, COL16A1, CXCL12, and MFAP4. Subcellular localization studies further confirmed that all four key genes were predominantly expressed in the extracellular space. ROC curve analysis revealed that only MFAP4 had an AUC value exceeding 0.9, indicating its high diagnostic value in BLCA. Ultimately, MFAP4 was selected for further research.

Microfibrillar associate protein 4 (MFAP4) belongs to the fibrinogen-binding protein superfamily and is a 36-kDa extracellular matrix (ECM) protein [42]. MFAP4 is highly expressed in the lungs, blood vessels, skin, and heart [43, 44], and it may serve as a biomarker for diseases, such as liver fibrosis, asthma, cardiovascular disease, dermatological conditions, and prostate cancer [45–49]. Several studies have suggested that MFAP4 may play a role in multiple human diseases. MFAP4 is highly expressed in contractile vascular smooth muscle cells, and it promotes the proliferation, migration, and monocyte chemotaxis of vascular smooth muscle cells via integrin αVβ3, thereby accelerating the repair of injured blood vessels [49]. Furthermore, knocking down MFAP4 enhances hepatocyte proliferation, accelerates liver regeneration, and reduces fibrosis, suggesting that MFAP4 has therapeutic potential for chronic liver disease [50]. Other researchers have reported that MFAP4 expression is significantly elevated in glioma tissues, where it promotes glioma cell proliferation, migration, and invasion. MFAP4 is also closely associated with poor disease prognosis, immune infiltration, and ferroptosis [51].

Further investigation is warranted to determine whether MFAP4, an extracellular matrix protein, is closely associated with key molecules involved in BLCA metastasis, metastatic structures, the tumor microenvironment, and the premetastatic microenvironment. Mechanical forces within the tumor microenvironment (such as solid stress and matrix stiffness) are key regulators of tumor invasion and act synergistically with ECM remodeling, cytoskeletal reorganization, and associated molecular signaling pathways. The matrix metalloproteinase (MMP) family serves as a core regulator of ECM remodeling, and the release of MMPs is dependent on the formation of actin-rich invasive pseudopodia. The assembly of invasive pseudopodia is closely linked to mechanically mediated cytoskeletal reorganization [52]. As a metastasis-specific structure of BLCA cells, the maturation of invasive pseudopodia relies on the targeted transport and secretion of MMPs (particularly MMP2). β1 integrin regulates MMP2 localization within invasive pseudopodia by mediating its endocytosis and recycling, processes that are also critical for bladder cancer cell invasion [53]. In the premetastatic microenvironment, MMPs (including MMP2) create conditions for tumor cell colonization by degrading the ECM and promoting vascular leakage [54].

MMP1 promotes tumor cell invasion and metastasis by degrading ECM components, such as type I collagen. MMP1 also modulates immune cell function through the CXCL16-CXCR6 and ANXA1-FPR3 signaling axes, thereby creating a protumor microenvironment [55]. MMP1 expression is further regulated by the JNK/C-Jun signaling pathway; the activation of this pathway upregulates MMP1 expression, thereby enhancing cellular migration and invasion capabilities [56]. MMP2 participates in BLCA invasion and distant metastasis by degrading basement membrane components. The regulatory mechanisms of MMP2 have been validated across multiple tumors. The ZNF24 zinc finger transcription factor inhibits MMP2 transcription by directly binding a specific 9-bp fragment of its promoter, thereby suppressing colorectal cancer growth and metastasis [57]. Conversely, the LRP1-SNRNP25 fusion gene activates the pJNK/37LRP/MMP2 signaling pathway to upregulate MMP2 expression and promote invasion and migration in osteosarcoma [58]. These findings suggest that the regulation of MMP2 expression is tumor-type specific but that its prometastatic function remains conserved. Previous studies have indicated that ECM proteins and MMP family members regulate tumor progression through direct or indirect interactions. It has been hypothesized that MFAP4 may participate in ECM degradation and remodeling by modulating the expression or activity of MMP1 and MMP2 (e.g., mimicking the transcriptional repression of ZNF24 [57], interfering with pJNK/37LRP-mediated MMP2 activation [58], or directly affecting their protein activity) to participate in ECM degradation and remodeling. Concurrently, MFAP4 may influence mechanical signaling in the tumor microenvironment by regulating the physical properties of the ECM (e.g., stiffness) [59], thereby inhibiting β1 integrin-mediated targeted transport of MMP2 to invasive pseudopodia [53] and reducing the migration and invasion capabilities of BLCA cells. Furthermore, MFAP4 may suppress MMP-mediated angiogenesis and immune suppression in the premetastatic microenvironment [54].

TGF-β, a key cytokine that regulates EMT and the tumor microenvironment, operates through a core signaling pathway mechanism. TGF-β induces SMAD2/3 phosphorylation to transmit protumor signals, thereby activating MMP expression and the EMT process [52]. Studies have confirmed that TGF-β regulates downstream molecules through interactions with target proteins. For example, RGS3 promotes SMAD2/3 phosphorylation by binding ARID3B, thereby enhancing the activity of the TGF-β signaling pathway and ultimately driving tumor cell proliferation and metastasis [52]. EMT processes are closely linked to tumor cell adaptation to the mechanical microenvironment. Mechanical forces induce EMT transitions by regulating the nuclear localization of EMT transcription factors, such as TWIST and SNAI [59]. In bladder cancer (BLCA), TGF-β modulates the actin cytoskeleton by inducing Transgelin expression, directly promoting invasive pseudopod assembly and tumor metastasis [53]. Furthermore, TGF-β contributes to the establishment of the premetastatic niche by inducing the recruitment of immunosuppressive cells (e.g., MDSCs and Tregs) and promoting inflammatory cytokine secretion [54]. MFAP4 has been implicated in TGF-β-mediated tissue remodeling in several diseases, such as liver fibrosis [43], suggesting that it may act through the TGF-β/SMAD2/3 signaling axis or by interfering with interactions between TGF-β and regulatory proteins, such as ARID3B [52]. Concurrently, MFAP4 inhibits the regulation of key invasion pseudopod molecules (e.g., cortactin and TKS5) [53] and immunosuppression in the TGF-β-mediated premetastatic microenvironment [54], thereby synergistically regulating BLCA EMT processes and tumor microenvironment homeostasis. MFAP4 may also influence mechanically mediated EMT activation by modulating ECM stiffness [59].

In the present study, MFAP4 overexpression inhibited BLCA cell proliferation, migration, and invasion. This phenotype is antagonistic to MMP1 and MMP2 overexpression, TGF-β pathway activation, mechanically induced invasion phenotypes, pseudopod-mediated invasion phenotypes, and premetastatic microenvironment-promoted metastasis phenotypes [52–59]. It is speculated that MFAP4 exerts its antitumor effects through multiple mechanisms by antagonizing the functions of prometastatic molecules (such as MMP1 and MMP2), interfering with the regulation of MMP1 by the JNK/C-Jun signaling pathway [56], block the activation of MMP2 by pJNK/37LRP [58], competing for binding to the MMP2 promoter region [57], and inhibiting TGF-β-mediated SMAD2/3 phosphorylation or the expression of its downstream target genes [52]. Concurrently, MFAP4 may suppress mechanically induced cytoskeletal reorganization and invasive pseudopod formation by stabilizing the ECM structure and regulating matrix stiffness [59]while also inhibiting inflammatory cytokine secretion and immune-suppressive cell recruitment in the premetastatic microenvironment [54]. The specific interaction mechanisms require further validation through subsequent co-IP, immunofluorescence colocalization, promoter binding assays, biomechanical-related tests (e.g., matrix stiffness measurement and cytoskeletal tension analysis), invasive pseudopod functional assays (e.g., gelatinase spectrum and invasive pseudopod imaging), and premetastatic microenvironment-related indicator tests (e.g., immune cell subset analysis and inflammatory factor detection).

The KM survival curve indicated a poor prognosis in the high-expression group, which does not negate the tumor-suppressing role of MFAP4 but rather reflects its complex function in bladder cancer and the heterogeneity of the tumor microenvironment. Some studies have suggested that low MFAP4 expression in bladder cancer tissues represents an early molecular feature of tumorigenesis, whereas high expression constitutes a compensatory response during tumor progression to intermediate-to-advanced stages. As bladder cancer progresses to late stages, the tumor microenvironment undergoes dramatic remodeling characterized by ECM fibrosis and increased matrix stiffness, promoting cell proliferation and migration. Consequently, high MFAP4 expression is a marker of disease progression and poor prognosis. Furthermore, elevated MFAP4 expression is closely associated with the immunosuppressive state of the bladder cancer tumor microenvironment, which is involved in immune cell infiltration and inflammatory regulation [43]. In bladder cancer, high MFAP4 expression may induce the accumulation of certain immunosuppressive cells, such as MDSCs and Tregs, within the tumor microenvironment, promoting tumor immune escape and consequently reducing patient survival rates. These findings also align with the GSEA results indicating the enrichment of regulatory T cells.

The present study evaluated the correlation of MFAP4 with three commonly used bladder cancer chemotherapy drugs—doxorubicin, gemcitabine, and cisplatin. Molecular docking analysis revealed a strong correlation between MFAP4 and these drugs, suggesting that the MFAP4 gene may serve as a promising therapeutic target for BLCA. These findings indicate that the MFAP4 gene may be strongly associated with BLCA development and may be a therapeutic target for BLCA.

In the present study, qPCR and WB analyses were performed to measure MFAP4 expression levels in UM-UC-3, 5637, T24, and normal SU-HUC-1 cells. The findings were consistent with those from the GEO dataset. Moreover, UM-UC-3 cells were transfected with a MFAP4 overexpression construct, and qPCR and WB analyses validated the transfection efficiency of the MFAP4 overexpression construct. Functional assays demonstrated that MFAP4 overexpression suppressed UM-UC-3 cell proliferation, migration, and invasion. Compared with those of the UM-UC-3 and NC groups, DOX treatment decreased the IC50 and cell proliferation rate of MFAP4-overexpressing UM-UC-3 cells, indicating that UM-UC-3 cells became more sensitive to DOX. These findings further confirmed that MFAP4 may serve as a novel biomarker and therapeutic target for bladder cancer.

MFAP4 holds significant potential for clinical translation as a biomarker. Its specific abnormal expression across multiple diseases enables its use as a marker for early screening. Furthermore, its expression levels correlate closely with the prognosis of various diseases, making it a crucial reference indicator for assessing prognosis. It also serves as a drug target to enhance the clinical efficacy of bladder cancer treatments.

The present study had several limitations. The present study relied primarily on online databases for screening and analysis, and it lacked the latest clinical samples and data. Consequently, the correlation between the expression of MFAP4 in clinical samples and its association with treatment efficacy could not be fully validated; thus, clinical translational application was not achieved. The present study did not optimize or validate the MFAP4 detection method, and it did not explore the MFAP4 expression levels or detection feasibility in bodily fluids, such as urine and serum. Consequently, direct experimental evidence for the development of noninvasive diagnostic kits is lacking. Furthermore, in vitro experiments to validate gene expression levels were performed using the UM-UC-3 cell line, which exhibited the most significant difference in expression, for MFAP4 overexpression and subsequent functional assays. Therefore, in future research, we will collect a large number of clinical samples and data, and we will conduct downstream pathway mechanism studies and animal experiments to further validate the present conclusions. Moreover, studies on multiresistant and doxorubicin-resistant cell lines and animal models are lacking, making it impossible to validate the specific role of MFAP4 in the in vivo development and metastasis of bladder cancer or its impact on mouse survival outcomes. This also hinders a comprehensive assessment of the efficacy of MFAP4 as a therapeutic target in vivo. Concurrently, cross-type studies involving different bladder cancer cell types need to be conducted.

Conclusion

Overall, the present study identified MFAP4 as a potential diagnostic biomarker and therapeutic target for BLCA through WGNCA and LASSO screening. MFAP4 inhibits the proliferation, migration, and invasion of BLCA cells while enhancing their sensitivity to doxorubicin chemotherapy.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

W.X.Z., H.L.S. and L.L.D. conceived this study; W.X.Z. and L.L.D. performed the bioinformatics analyses; W.X.Z. and M.Z.L. conducted the cellular experiments and data analysis. W.X.Z. and H.L.S. jointly drafted and revised the manuscript. All individuals meeting the criteria for authorship are listed as authors. All the authors participated in reviewing the manuscript.

Funding

Postgraduate Research & Practice Innovation Program of Jiangsu Province (SJCX22_1880); Research Project of Changzhou Medical Center of Nanjing Medical University (CMC2024PY30).

Data availability

The gene expression datasets analyzed in this study are available from the Gene Expression Omnibus (GEO) database under the accession numbers GSE7476, GSE13507, and GSE52519 (https://www.ncbi.nlm.nih.gov/geo/). Subcellular localization information for key genes was obtained from the GeneCards database (https://www.genecards.org/). Protein‒protein interaction networks were constructed and analyzed using the STRING database (https://string-db.org/), with an interaction score threshold of >0.3. The datasets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable. All data used in this study are publicly accessible, thus participant consent is not required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.